Table 3. List of model dimension, number of layers, code language, number of subroutines, number of lines, time step, and recommended hardware

Model Name #Dimension #Layering Language #Subroutines #Lines Time Step Hardware
Australian - - Statistical regression model run on Excel spreadsheet - - Daily Any
BAIM 1D 3 FORTRAN 29 5000 1 hour or less -
BASE 1D 1 FORTRAN 4 357 including comments Host model determined - 20-30 mins usually anything - runs on SGi, PCs, Suns, CRAY etc.
BATS 1D 1 mass and 2 thermal layers FORTRAN 77 7 1019 5 minutes to 3 hours, mostly 20-30 minutes PC or UNIX computers
CLASS 1D 1 FORTRAN Many Numerous 30 minutes (specified by user) PC or UNIX computers
CROCUS 1D 50 (variable depth) FORTRAN 77 23 1800 effective lines 900 seconds most unix workstations
DARSSM 1D 4 FORTRAN 7 (incl. soil model) 3000 20 to 30 min -
DHSVM 1D 2 C 5 1550 24 min to 1 hour Unix workstations (SUN, HP) and Pentium class PC's
ECHAM (Duemenil) 1D 1 FORTRAN 1 - not fixed (GCM: 24; input data: 1 h to 3h) Cray
ECHAM (Loth) 1D 2-5 FORTRAN 7 - not fixed (GCM: 24; input data: 1 h to 3h) Cray
GFDL Snow Model 1D 1 FORTRAN 77 - - Same as GCM -
Hadley Centre/UKMO GCM LSM 1D 4 soil layers. Snowpack is included in surface soil layer. FORTRAN 20 5000 30 minutes in climate GCM -
HBV 1D 1 Most in FORTRAN, newest in C++ 4 major parts Depending of programming strategy (a few to thousands) one day DOS on a ordinary PC (for instance, a 386), newer versions demand Windows-95, NT or unix
IAP94 1D 3 FORTRAN 77 10 - 30 min to 1 hour Convex C3-120, SGI workstation, DEC workstation
IHACRES Snow Model 2-3D 1 FORTRAN 77 10 2000 Daily SUN. Probably Pentium is OK, but with large hard disk
INM 1D The model treats the snowpack as a unique layer, but the freezing depth subdivides the snowpack into a dry surface layer and a humid layer FORTRAN 90 9 subroutines. The main one treats the internal evolution 300 effective line flexible according to input data (e.g. 1 hour) -
ISBA 1D - FORTRAN 90 4 subroutines The snow code is embedded in the ISBA land-surface scheme, whose "off-line" version includes about 1350 lines between 5 and 30 minutes PC or work station
ISBA-ES 1D 3 F77 and 90 approx 25 subroutines F90 2326 total lines: of which less than approximately 50% are actual executable lines of code: the rest are comments. between 5 and 30 minutes (time scheme is fully implicit) The code has been compiled and run on: F77: UNIX HP workstation, Linux PC 199 GHz Pentium F90: UNIX HP workstation, CRAY ymp, Fujitzu (similar to CRAY) (Code in F77 is scalar, F90 has 2D or 3D options)
LSMCS 1D 3L FORTRAN 28 subroutines and functions more than 2000 lines 1800 seconds any type of hardware
MAPS/RUC SVS 1D 1L F77 with extensions Three subroutine Approximately 500 5-min time step is used in MAPS/RUC forecast model No special requirements.
MAR 1D ML F77 17 4898 (2382 effective lines) 6 minutes Any unix workstation (model has been used so far on DEC and HP workstations)
Mosaic 1D 1L FORTRAN There are specific lines regarding snow in several of the LSM subroutines - Between 5 and 20 minutes many different systems
MRI-CGCM GHM 1D 1L Fortran 77 5 1052 1 hour super computer; Hitachi s-3800
MU-SNW One dimensional, but includes information about the sub grid scale topography. 1L FORTRAN About 5 About 300 About 20 min very portable
NCEP/OH/OSU CAPS 1D 1L FORTRAN 90 one primary subroutine and significant parts of 3-4 other subroutines about 250 10-60 minutes UNIX workstation or UNIX mainframe (e.g. CRAY)
NCEP/OH/OSU CAPS 1D 1L (snow), 4L (frozen soil) FORTRAN 9 about 700 Snow-frozen ground model time step depends on a land surface required time step that is usually 30 minutes UNIX workstation
NWSRFS SNOW-17 - - Fortran - - - -
RAMS Snow Component 1-D 1L Fortran Most code contained within two subroutines, but interspersed with code for soil temperature determination and soil hydrology Approximately 30 dedicated to snow 90 seconds, dictated by atmospheric component of the model RISC-based workstation. I use an HP 735 and will soon acquire an HP C180
RGM Horizontally distributed, vertically lumped One snowpack layer Interactive Data Language (IDL) Approx. 20 procedures plus a programming library of hundreds of procedures and functions for a large variety of purposes (e.g. GIS, RDBMS, mapping, math ...) More than 1000 lines without the library code (rapidly increasing) User-selectable (typically 1 hour) Each system supporting IDL (PC, VMS, Unix, Mac, SG)
SEMS One-dimensional energy balances, 3-dimensional snow distributions, that evolve in time. Multi-layers, layers defined depending on application; usually identified by a continuous precipitation event. FORTRAN about 55 about 4000 Determined by individual application; typically 30 minutes to a day run on UNIX workstations.
SHAW One-dimensional profile model The model will break the snow pack into layers; the thickness of the layers will depend on its depth within the snow pack. The surface layers will be approximately 2 cm thick and the layer thickness will increase with depth.Layers at a depth of 2 m will be about 10 cm thick. FORTRAN 77 total subroutines; approximately 10 dealing specifically with snow 9000 total linesof code; approximately 950 directly related to snow Hourly or daily None recommended; the model is very transportable between hardware
SLURP one dimensional applied to areas of common land cover one layer FORTRAN 90 integrated in larger model 50? daily Pentium
SNAP One-dimensional Currenly "effective pack"= one layer, but a multiple layer version is currently under development. FORTRAN 12 approx 2500 Since it uses a new analytical formulation for water movement through the pack (eliminating the need for matrix computations), there is no restriction on the time step. The time step is dictated by the frequency of the input data. Hourly input data is currently what we are using 486 pc or better
SNOWPACK 1D (operational), 2D (research) Finite Element layers, with each snowfall new layers/elements are added C Many approx. 20000 including comments Typically 15 minutes UNIX/PC
SNTHERM 1D Unlimited and arbitrary - changes during run. FORTRAN 77 66 10993 Model uses adaptive time step which adjusts to meet convergence criteria (much smaller step during water flow). User inputs max and min of range. Max is equal to or less than time step of input meteorological data PC or UNIX
SNTHERM (Davis) 1D Unlimited and arbitrary - changes during run FORTRAN - SNTHERM; spatial distribution => UNIX shell and C 75 10993 Model uses adaptive time step which adjusts to meet convergence criteria (much smaller step during water flow). User inputs max and min of range. Max is equal to or less than time step of input meteorological data SNTHERM - PC, MAC or UNIX spatial distribution - UNIX
SNTHERM.ver4 One-dimensional, with three D radiative transfer ca. 20 SNTHERM: FORTRAN , calls shellscript which calculates parameters for DISORT radiative transfer program (Stamnes et al. 1988); UNIX Extra; 2 + DISORT (many) - variable. Hourly data UNIX
SOIL 1D 1L FORTRAN The whole SOIL-model has about 77 subroutines, one of which is the snow-subroutine. ca. 100 It is set by the user, often hourly or daily inputs. PC
SPONSOR 1D 2 (at both boundaries of snow cover) FORTRAN 4 subroutines for the snow block only and 3 subroutines dealing with the snow characteristics. About 400-450 (some operators deal not only with snow) 30 to 60 minutes No specific recommendations.
SPS (CAPS-LLNL) single snow layer on top of multiple soil layer 1L Fortran - - 5 min - hours (usually 5 min for coupled model, column version was run up to 3-hr steps). workstations - CRAY
SSiB 1D 1L FORTRAN 3 about 100 lines from 2 min to 1 hour no
TSCM1 1D 1L FORTRAN and BASIC One 150 One hour PC
TSCMM 1D the thickness is 2cm number of layers depends on snow depth FORTRAN 14 600 200 sec PC or work-station
UEB 1D 1L FORTRAN About 10 About 600 Flexible but longer than 6 hours not recommended Any with a Fortran compiler - Has been successfully run on Sun Solaris, IBM RS6000 and Intel 486 and Pentium hardware.

(Prepared by Zong-Liang Yang and Guo-Yue Niu)

-- Last updated Fri Oct 8 12:47:54 MST 1999 by Zong-Liang Yang.
For questions and comments, please contact Zong-Liang Yang